Background & Motivation

• Elite perceptions and responsiveness to public threats (e.g. COVID-19, climate change etc.)

• Delays in elite response to COVID-19 likely cost lives

• Misperception, motivated reasoning, and polarization

• Exogenous shocks and political attitudes/behaviour

  • Natural disasters and climate change attitudes (Baccini and Leemann 2021; Kalatzi Pantera, Böhmelt, and Bakaki 2023)
  • Daughters and feminist political attitudes (Glynn and Sen 2015)
  • Support for war and children’s military service (McGuirk, Hilger, and Miller 2023)

Measurement & Descriptive Results

• Outcome: Number of messages about COVID-19 (at t) that are oppositional as a proportion of the total number of COVID related messages (at t) (n = 82,870)

Estimation

• Treatment variable = the day on which a legislator is infected with COVID-19 and the two weeks following

• The treatment switches on and off for each legislator at different times

• Legislators can be infected multiple times

• We treat the outcomes of infected legislators as missing values and use low-rank matrix completion to estimate the counterfactual (Athey et al. 2021; Liu, Wang, and Xu 2022)

Results

• Treatment effect: COVID-19 infection reduces opposition to COVID-19 policies by about 10% (95% CI: ~2%-17%)

How Long does the Effect Last?

Not Long

Heterogenous Effects?

• Does the effect vary by party?
  • Somewhat, with suggestive evidence that Democrats are more responsive, though not statistically differentiable
• Does the effect vary by gender?
  • More likely - infection affects men’s expressed opposition slightly more so than women’s
  • This is likely due to gender differences in opposition to COVID-19 policies in general

Robustness

• Legislators do not tweet less about COVID-19 when infected with COVID-19
  • In fact, they tend to tweet more, but not in a statistically measurable way
• Effects are still identifiable without controls for time-varying confounders (e.g. COVID-19 cases, deaths)
• Effects are still identifiable when using a static treatment variable
  • Slightly larger, statistically distinguishable (~30%)

Conclusion

• Main takeaway: COVID-19 infection reduces opposition to COVID-19 policies on social media
  • This effect is short-lived, lasting only about two weeks
  • Hard to disentangle whether the effect is due to risk realization or strategic avoidance of the issue
• Implications:
  • Elites are not immune to the effects of exogenous shocks
  • Both good and bad news for elites’ responsiveness to public threats
    • Good news: - responsiveness likely to be true for other issues as well (e.g. climate change, natural disasters, etc.)
    • Bad news: the effects are short-lived and may not be enough to overcome polarization and motivated reasoning